\documentclass[11pt]{article}
\usepackage[left=1in,right=1in,top=1in,bottom=1in,letterpaper]{geometry} 
%\usepackage{booktabs}
\usepackage{dcolumn}
\usepackage{natbib}
\usepackage{setspace}
\onehalfspacing
\usepackage{fullpage}
\usepackage{paralist}
\usepackage{rotating}
\usepackage{hyperref}
\usepackage{pdfpages}
\renewcommand*\thesubsection{\arabic{subsection}}


\begin{document}
\begin{center}
{\bf \Large \onehalfspacing Online Appendix for ``Americans' Belief in Linked Fate"} \\ 
June 2015
%\vspace{6pt}
\end{center}

\section{Mechanical Turk Survey}
As described in the paper, we used Amazon's Mechanical Turk service to recruit participants for a second round of data collection in April and May 2015, using a two-wave sampling process to obtain respondents.  This section presents analysis of responses to our main survey, sent to 632 people in May 2015.

\subsection*{Demographics}  
We collected 449 responses, with 86 Black, 88 Hispanic, 135 Asian, and 140 White respondents. A slight majority of respondents (266/449) were male.  As is common on Mechanical Turk, the sample skews heavily Democratic (80\% of the sample reported being Democratic or leaning towards the Democratic party).  Respondents were less religious than the American public as a whole, with 221 reporting no religious affiliations (of the remainder, 185 observed some form of Christianity and 43 held other faiths).  The vast majority of respondents identified as either middle class (228) or working class (136), with relatively few identifying as lower class (77) or upper class (8).  

The main purpose of this follow-up study was not to find another representative sample, but to test some survey items that might capture the same thing as the linked fate item.  As these were fairly universal concepts and relationships, we were comfortable running this survey on a convenience sample. In order to keep the survey short, we did not ask many demographic questions beyond those needed to construct our Linked Fate items and outcome measures. However, we can look to other large MTurk surveys as a guide to the demographics of the sample pool.  For example, we did not collect measures of education, so we cannot explore whether Black respondents to our sample are especially highly-educated(which has been associated with higher levels of both linked fate and political engagement).  However, two major surveys characterizing MTurk samples are useful here.  In our own analysis of the \href{https://dataverse.harvard.edu/dataset.xhtml?persistentId=hdl:1902.1/17220}{replication data from Berinsky, Huber, and Lenz (2012)}, we found that Black respondents to their survey reported an average of 13.4 years of education.  For reference, this is higher than the mean education reported by overall sample of respondents  (all races) to the 2008 Current Population Survey: it is quite high.  Similarly, Black respondents to the recent MTurk surveys described in \href{http://scholar.harvard.edu/dtingley/publications/who-are-these-people-evaluating-demographic-characteristics-and-political-0}{a working paper by Huff \& Tingley (2015)} reported very high levels of education, with only 17.5\% of Black respondents reporting that they had a high school education or below (personal communication, Connor Huff, June 2015).  We suspect that such high levels of education in this sample could be the reason that we observe somewhat higher levels of Black racial linked fate here than in LFSS.  

\clearpage
\subsection*{Replicating LFSS figures/analysis with MTurk data} 
\begin{figure}[ht!]
\begin{centering}
\includegraphics[scale=.7]{LF_levels.pdf}
\caption{Levels of Different Types of Linked Fate in the Mechanical Turk Sample} 
\end{centering}
\end{figure}

\clearpage

\begin{figure}[ht!]
\begin{centering}
\includegraphics[scale=1]{LFbyrace_fig1rep.pdf} 
\caption{Levels of Linked Fate by Race, Mechanical Turk Sample} 
\end{centering}
\end{figure}

\clearpage

\begin{figure}[ht!]
\begin{centering}
\includegraphics[scale=.7]{sociologicaldiffs_june2014_Mturk.pdf} 
\caption{Linked Fate by Group Status, Mechanical Turk Sample} 
\end{centering}
\end{figure}
\clearpage

\input{"LFrace_politicization.tex"}

\input{"LFclass_politicization.tex"}

\input{"LFgender_politicization.tex"}

\input{"LFreligion_politicization.tex"}


\input{"LFrace_politicization_black1.tex"}


\clearpage
\section{Mechanical Turk Survey Instrument 1}
This document presents the text of the demographic survey used in the first wave of our Mechanical Turk data collection, as discussed in the paper.
\includepdf[pages=1-4]{wave1survey_blinded.pdf}



\clearpage
\section{Mechanical Turk Survey Instrument 2}
The next document is the full survey instrument used for the main Mechanical Turk survey discussed in the paper.  Note that many variations on the linked fate item appear in the survey, but each respondent was only asked two variations.\footnote{Due to an error in survey flow, some people who received the racial linked fate item at the beginning of the survey were asked it again at the end, rather than receiving one of the three other linked fate items.  This affected only the last question of the survey; it reduces our power somewhat, as it means fewer people answered those items, but it should not change the answers to any other survey questions. }  

\includepdf[pages=1-15]{survey2_blinded.pdf}


\end{document}